For the data-driven enterprise, visualization isn’t just a tool in the toolbox, argues Scott Berinato — it’s a core skill in the fight for competitive advantage.
Scott Berinato, author of Good Charts and the Good Charts Workbook, and a senior editor at Harvard Business Review, says complex charts and visualizations arose as decision-making tools to help railroads and shipping companies makes sense out of data overload.
Today’s businesses have exponentially more data, but for many, visualization remains an underdeveloped and utilized capability.
Berinato explains — with the help of a few charts, naturally — how a culture of data visualization can help create a truly data-driven enterprise.
Connected Futures: In your books you talk about how data visualization really gained traction as a way to see, understand, and fix complex business processes.
Scott Berinato: Data visualization was really a result of the increasing complexity of organizations in the late industrial revolution. The first people to visualize at scale were the large industrials like the railroad companies, and large municipal functions like ports and utilities. These were organizations that were looking for ways to understand and manage very new and very complex systems at scale.
And they were doing some really cool stuff. Double-axis timelines; electrical loads charted by the hour and broken down by source of demand; analysis of vacation schedules’ effect on productivity. It’s impressive.
In Willard Brinton’s brilliant 1914 book Graphic Methods for Presenting Facts, he shows dozens of these incredible charts, like this one:
They were made by hand. They were core decision-making tools, important to running these big businesses.
When you read Brinton you realize a couple of things. One, not much is new in charting. They were doing stacked area and slope graphs and floating bars then, too. And two, charting emerged because complexity emerged. It was a necessary invention.
Today companies need and want to be “data-driven”. But cognitive bias really affects how humans interpret data — for example, people are pretty bad at managing risk, even when you show them data. Can visualization play a role in combating cognitive bias?
It can play a role in both driving bias, and removing it.
I have a slide for my talks that shows a sharp knife on it and says “Every Chart Is a Manipulation.” Data scientists hate this, but it’s accurate.
There is not one ‘true’ chart [for any given data set]. All charts are a collection of decisions — conscious and subconscious — about what to show, and not show, and how to show it. Just changing shapes (not data) changes stories. In one parlor trick, I take a chart that seems to show swelling immigration levels, and without adding or taking away a single data point, I flip the meaning entirely in five steps.
What we need to understand above all is that people do not read statistics. They do not see facts. They form ideas. It’s the Law of Prägnanz , a gestalt idea that basically says our minds pick the fastest, best explanation. As Kurt Koffka said: “The whole is other than the sum of the parts.” For example:
We see four things here: Three circles and a triangle. We can’t not see that. But in fact the “data” is just three arced wedges. The same thing happens when we look at a chart. We see steep lines and think “big change,” not “the value rose 28%.” We see flat lines and think nothing’s happening, even if a flat line describes, say, climate change.
We have to understand these biases and be prepared to defend our visuals and call out misleading visuals. Effectively it’s about data visual literacy, which can be taught. I put the knife on the slide to show that just because something can be used poorly, or illicitly, doesn’t mean it shouldn’t be used. Because it can also be used skillfully to great effect.
Any organization with a lot of humans inevitably has a lot of disagreements. Can you make a case for dataviz helping organizations work better together? Or is that a bridge too far?
I’d go further and say if a company doesn’t use visualization to make sense of their data, they’ll lose. It’s a competitive imperative. Data is too complex to do anything with without good visualization.
And now that some companies are getting much better, they are seeing threats and opportunities those who don’t adapt won’t see. Daryl Morey, general manager of the Houston Rockets, said to me “We’re excited when we can use it right and others use it wrong.”
Example: Tesla uses visualizations to assess how people actually drive their cars. It’s so much data you can’t do anything with it in raw form but with visualization, they can adjust how they engineer their cars based on actual behavior rather than ideal behavior.
</section><blockquote>When you present to the board, you invest in a haircut, a suit, speaking lessons. Then you show them charts they can't understand.</blockquote> </div> <p class="quote-author">- Scott Berinato</p> </div> <section>But it takes an investment and a cultural mindset.
Even on the individual level I hear people say, “It just seems like a lot of work” and I say, ‘You present to the board, you invest in a haircut, a suit, speaking lessons. You invest time in practicing. Then you stand up there and show them charts they can’t understand and end up using all your time trying to explain to them what they should see.
So maybe the investment in good charts is worth it.”
What are two or three real-world examples of data visualization that you have found most surprising or compelling?
I’ve found some of the politically motivated visualization — showing gun deaths or climate change, for example — to be compelling. Here’s an example.
Most people tend to react to things that are beautiful or that say something clearly that they’ve felt, but I also tend to be enamored with people who are exploring with viz. Here’s an example. My colleague casually bemoaned the fact that there were so few 70-degree days in Boston, and we asked casually ‘is that true?’ And we explored. We discovered some interesting things. First, there are many more days without highs in the 70s than with in Boston:
But a histogram reveals that it is, by a narrow margin, the most common 10-degree range of high temperature:
Then we just started thinking about what we could do with it. We could look at the range of high temps each month, which showed the greatest variance is really the dead of winter:
Once we did it we started mapping out how you might adapt this for, say a travel site. (“I want to go to Miami. What is the weather like during my trip…”)
So that kind of exploratory thing really gets me excited.
Where do organizations most often go wrong in building a culture or habit of effective visualization?
They build their teams poorly. They focus on getting as much data talent as they can and leave the design and storytelling to the end and don’t make that part of their core team.
Or they leave the data analysis to the data scientists. Those are different talents and many people smarter than me will tell you the bests analysts are liberal arts majors, not coders.
I recently wrote about this for Harvard Business Review: https://hbr.org/2019/01/data-science-and-the-art-of-persuasion.
So what should a CEO do about all this? Require training across the whole company?
I certainly advocate for data scientists to learn some basic design and for designers/storytellers to absolutely learn Stats 101. Empathy is powerful. They should follow the advice in the above article about building teams that have all the talents you need, not just the coders best at maths and algos. These teams should sit together and all be involved along the course of a project.
They should invest in templatizing; It doesn’t cost that much and makes the process of getting good output for basic chart types much easier, so people don’t have to choose colors or bar widths.
They should run workshops that teach people how to think through their challenge and sketch good solutions.
And they should reward good viz. For fun when I do workshops and have people present, we give people rewards for showing visuals without explaining how they’re structured. Inevitably, presentation is part of what makes a good chart, too. When you put a chart up you shouldn’t have to say “as you can see here” or “the x axis is quarters and the y is revenues and the gray line is…” etc. I see that. Why are you telling me?
And if you’re telling me because I can’t see all that, then we need to fix that chart.
What else should people ask about data visualization?
Some people ask about VR and AR. I’m in wait-and-see mode on that. I have reservations about “putting people in the chart.”
Lots of people ask me “What tools should I use?” and the answer is unsatisfyingly complex. No one tool does everything well. All the tools do some things well. The more powerful the tool, the steeper the learning curve. We’re still waiting for that “word processor of visualization” tool. I’ve seen some promising prototypes but nothing real yet.
Also, paper and colored pencils are great, or an iPad. Most of my charts are close to realized in sketch form before we get to the actual creation. Sketching works.
Did you like this article?